1446 MONTHLY WEATHER REVIEW VOLUME 140

Tornado Climatology of

JENNI RAUHALA Finnish Meteorological Institute, Helsinki, Finland

HAROLD E. BROOKS NOAA/National Severe Storms Laboratory, Norman, Oklahoma

DAVID M. SCHULTZ Centre for Atmospheric Science, School for Earth, Atmospheric and Environmental Sciences, University of Manchester, Manchester, United Kingdom, and Division of Atmospheric Science, Department of Physics, University of Helsinki, and Finnish Meteorological Institute, Helsinki, Finland

(Manuscript received 31 July 2011, in final form 8 November 2011)

ABSTRACT

A tornado climatology for Finland is constructed from 1796 to 2007. The climatology consists of two datasets. A historical dataset (1796–1996) is largely constructed from newspaper archives and other historical archives and datasets, and a recent dataset (1997–2007) is largely constructed from eyewitness accounts sent to the Finnish Meteorological Institute and news reports. This article describes the process of collecting and evaluating possible tornado reports. Altogether, 298 Finnish tornado cases compose the climatology: 129 from the historical dataset and 169 from the recent dataset. An annual average of 14 tornado cases occur in Finland (1997–2007). A case with a significant tornado (F2 or stronger) occurs in our database on average every other year, composing 14% of all tornado cases. All documented tornadoes in Finland have occurred between April and November. As in the neighboring countries in northern Europe, July and August are the months with the maximum frequency of tornado cases, coincident with the highest lightning occurrence both over land and sea. Waterspouts tend to be favored later in the summer, peaking in August. The peak month for significant tornadoes is August. The diurnal peak for tornado cases is 1700–1859 local time.

1. Introduction 2005), a case study of a severe thunderstorm outbreak (Punkka et al. 2006), micrometeorological measure- As late as the mid-1990s, Finnish meteorologists gen- ments of a microburst (Ja¨rvi et al. 2007), a severe hail erally assumed that severe convective storms or torna- climatology (Tuovinen et al. 2009), and several case studies does did not occur in Finland and, if they occurred, they of tornadoes (e.g., Teittinen et al. 2006; Teittinen and were rare. Tornado reports were not collected, and no Ma¨kela¨ 2008; Outinen and Teittinen 2008; Rauhala and research on severe convective storms was published Punkka 2008). Yet, no climatology of tornadoes exists from the 1960s until recently. However, since 1997, severe for Finland. thunderstorms, and especially tornadoes, have received Historically, in Italy and , scientific papers on a lot of media attention in Finland. Thus, Finnish mete- tornado cases were published by the seventeenth cen- orologists have started to appreciate the occurrence and tury (Peterson 1982). The first research covering all of threat of severe weather. This appreciation has led to Europe was Wegener (1917). Estonia, the southern neigh- several studies on severe weather in Finland: a climatol- bor of Finland, has a long history in tornado research. Be- ogy of mesoscale convective systems (Punkka and Bister tween the wars, tornado research was done by Johannes Letzmann (Peterson 1992, 1995) and recently by Tooming et al. (1995) and Tarand (1995). In , several case Corresponding author address: Jenni Rauhala, Finnish Meteo- rological Institute, Erik Palme´nin Aukio 1, P.O. Box 503, FI-00101, studies have been performed (Peterson 1998). Helsinki, Finland. More recently, European tornado climatologies have E-mail: jenni.rauhala@fmi.fi been published for Ireland (Tyrrell 2003), the United

DOI: 10.1175/MWR-D-11-00196.1

Ó 2012 American Meteorological Society Unauthenticated | Downloaded 10/01/21 07:20 AM UTC MAY 2012 R AUHALA ET AL. 1447

Kingdom (Elsom and Meaden 1982; Reynolds 1999; with the ground, either pendant from a cumuliform cloud Holden and Wright 2004), Lithuania (Marcinoniene 2003), or underneath a cumuliform cloud, and often (but not and the former Soviet Union (Peterson 2000). The always) visible as a funnel cloud.’’ Forbes and Wakimoto climatology of French tornadoes has been studied by (1983) suggested that a vortex would be classified as Dessens and Snow (1989, 1993) and Paul (2001). Tornado a tornado if it were strong enough to cause at least F0 statistics for Germany and Austria have been pub- (Fujita 1981) damage. These definitions were adopted in lished by Dotzek (2001) and Holzer (2001), respectively. the United States where all tornadoes have been classi- Climatologies for the by Setva´ketal. fied by the Fujita scale (F scale), even if there were no (2003) and for Hungary by Szilard (2007) have been damage. Since 2007, tornado intensity has been rated constructed. In southern Europe, tornado climatologies in the United States with the Enhanced Fujita scale (EF have been documented in Greece (Sioutas 2003, 2011), scale; Doswell et al. 2009). On the other hand, water- the Balearic Islands (Gaya´ et al. 2001), Spain (Gaya´ 2011), spouts that do not hit land are not classified as torna- Portugal (Leitao 2003), and Italy (Giaiotti et al. 2007). does in the United States. This definition would be Despite a long history of observing tornadoes, not all a serious restriction for Finland, where almost 188 000 European countries have a database of severe weather lakes cover the landscape. In this work, the following reports. The European Severe Weather Database tornado definition is used: A tornado is a vortex be- (Brooks and Dotzek 2008; Dotzek et al. 2009) has im- tweenacloudandthelandorwatersurface,inwhichthe proved the collection of severe weather reports from connection between the cloud and surface is visible, or around Europe. Because many European national hydro- the vortex is strong enough to cause at least F0 damage. meteorological services are actively developing their This definition allows all waterspouts to be included in severe thunderstorm forecast and warning services the definition of a tornado. This is consistent with the (Rauhala and Schultz 2009), many countries have ac- Glossary of Meteorology (Glickman 2000) definition, knowledged that understanding the local climatology is which defines a waterspout as ‘‘any tornado over a body important for forecasting. of water.’’ Also, tornadoes over land that do not cause The purpose of this article is to create a tornado clima- damage, but have a visible connection between ground tology for Finland. The definition of a tornado, methods and the cloud base, are included. to collect reports, and the credibility evaluation process In this article, we will also make use of the concept of are discussed in section 2. Sections 3–4 summarize the a tornado case. In one tornado case, there might be many geographical and intensity distributions of tornado cases, tornadoes. For example, several tornadoes might occur and sections 5–6 summarize the monthly and diurnal dis- near each other (e.g., within the same storm, along the tributions. Section 7 concludes this article. same boundary, etc.), but this would be classified as one tornado case. This concept improves the database be- cause, in several waterspout cases, for example, the 2. Data exact number, location, or timing of each individual This section describes the development of the Finnish tornado is not known. Because waterspouts often occur tornado climatology. Because 10% of Finland is covered in groups of 5–20 single tornadoes, if each were recorded by freshwater and because Finland has a relatively low as an individual event, the monthly, diurnal, intensity, and population density, we take care in defining the criteria geographical distribution of waterspouts would dominate for tornadoes and waterspouts. Section 2a addresses this the database of all tornadoes. issue. Then, having described our definition for a tor- On the other hand, if tornadoes are known to occur in nado, the climatology is constructed in two steps. The separate storms, or the starting points of the successive first step was to collect historical reports of tornadoes tornadoes can be resolved, they are considered to be from archived sources such as newspapers. The second separate cases. There is often not enough information to step was to collect recent reports of tornadoes in near– split a report into several cases. Particularly in the his- real time through the Finnish Meteorological Institute torical dataset, the path length would often suggest a (FMI). Section 2b discusses the different methods em- series of tornadoes, but the event is still recorded as ployed to compile and evaluate tornado reports in these one case due to the lack of detailed information on two datasets. Finally, section 2c describes how the tor- damage tracks. The problem of distinguishing between nado intensities are assessed. long-track tornadoes and a series of short-track torna- does is discussed by Doswell and Burgess (1988). The a. Tornado definition and criteria starting point of the first tornado path in each event The Glossary of Meteorology (Glickman 2000) defines characterizes the case on geographical maps. In several a tornado as ‘‘a violently rotating column of air, in contact instances, the same tornado moves over both water and

Unauthenticated | Downloaded 10/01/21 07:20 AM UTC 1448 MONTHLY WEATHER REVIEW VOLUME 140 land surfaces. If a tornado is first observed over land, it is TABLE 1. Credibility categories of tornado reports. The report is classified as a tornado over land. If it is first observed attributed to the highest class where any of the criteria are satisfied. over water, it is classified as a tornado over water. From Category Criteria here on, ‘‘tornado over water’’ and ‘‘waterspout’’ will Confirmed A photograph or videotape of a tornado be used interchangeably. A tornado day in this article is Damage survey indicates tornado damage defined as a day with at least one tornado case. A tor- Probable Credible eyewitness observation of a tornado nado report is the original tornado observation (from Credible eyewitness report of typical tornado media, public, etc.); several tornado reports may exist damage for each tornado case. A photograph of a typical tornado damage b. Collecting tornado reports Possible No eyewitnesses Cause of the damage is not confirmed by Because the methods of collection, evaluation, and the observations of an eyewitness classification of tornado reports have changed over time, the tornado climatology of Finland comprises two data- sets. The historical dataset covers 1796–1996 when no event was too limited. This process resulted in 129 tor- systematic tornado documentation was maintained at nado cases. FMI. The recent dataset covers 1997–2007 when FMI In the recent 1997–2007 dataset, most of the pre- has been actively collecting information on tornadoes liminary reports were obtained from the general public. in Finland. The main source for these reports was through a tornado The tornado reports in the 1796–1996 historical dataset observation report form placed on the FMI Web site were collected through five different approaches. First, (http://www.fmi.fi) since 1998. In addition, preliminary most of the reports in the historical dataset come from reports were received from the general public by phone old newspaper articles. The oldest articles were found and by e-mail. Reports from news media were collected in the microfilm archives and digital collections of ma- also, including related newspaper articles and reporter jor newspapers ( Allma¨nna Tidning, Ilmarinen, information. Data from different sources were gathered Keski-Suomi, Porin Kaupungin Sanomia, Sanomia Turusta, into tornado cases and, in all except a few cases, eye- Tapio, Vesta Nyland, A˚ bo Tidningar,andA˚ bo under- witnesses were interviewed. Photos and videos of the ra¨ttelser) in the National Library of Finland. Second, tornado and its damage were collected and, in some FMI’s archives contained an unsystematic collection cases, a damage survey was performed. of weather-related newspaper clips and of FMI-related The credibility of a report was evaluated in the recent articles from the 1990s. The more recent newspaper ar- 1997–2007 dataset based on the observed information ticles were from Aamulehti, Borga˚bladet, Etela¨-Saimaa, available on the event. Each case was categorized as Etela¨-Suomi, Helsingin Sanomat, Iitinseutu, Ilta Sanomat, confirmed, probable, or possible (Table 1). Similar defi- Kainuun Sanomat, Kaleva, Karjalainen, Keskisuomalainen, nitions have been used in Canada to evaluate tornado Kuorevesi-Ma¨ntta¨-Vilppula, Lapin Kansa, Maakansa, reports (Newark 1984). Only confirmed and probable Savon Sanomat, Suomenmaa, Suur-Keuruun Sanomat, tornado cases were accepted into the recent dataset, Syd-O¨ sterbotten, Turun Sanomat, Uusi Aura, and Uusi which has now become the official tornado dataset for Suomi. Altogether, the archived and more recent news- Finland. Radar pictures were also studied for the tornado paper articles resulted in 98 tornado cases. Third, 17 cases of the last 8 yr (2000–07). If there was no radar echo tornado cases were obtained from the general public during or after a reported tornado, and satellite pictures during the 1970s and 1980s and were documented in did not show cloudiness, the case was not included in the the forecasters’ notebook—an unsystematic collection recent dataset. This process resulted in 169 tornado cases. of reports of strange phenomena received from the The historical dataset is undoubtedly incomplete, par- general public. Fourth, 12 cases, mainly from the 1930s, ticularly the records of weak tornado cases, as indicated come from published meteorological descriptions by the change in number of tornado cases per decade (Kera¨nen 1930; Simojoki 1931, 1935; Angervo 1949; Rossi (Fig. 1a). Before the 1930s, there are only a few, if any, and Franssila 1960). Fifth, two tornado cases are from known reports per decade. In the 1930s, 30 tornado cases Wegener’s (1917) book on European tornadoes. Be- (10 of which had a F2 or greater tornado; Table 2) af- cause of the lack of detailed information on the tornado fected Finland, causing two deaths, several injuries, and cases,thesehistoricaldatadidnotgothroughanex- vast material damage, which awoke the interest of me- tensive quality control and most of the reported tor- teorologists during that period. Both the attention paid nado cases were included in the statistics. Few cases to tornadoes and the availability of documented reports were rejected because the available information on the are reflected in the statistics with a larger number of

Unauthenticated | Downloaded 10/01/21 07:20 AM UTC MAY 2012 R AUHALA ET AL. 1449

discussed by Doswell and Burgess (1988), although sys- tematic biases may occur. The information available on events in the historical dataset is often so limited that an accurate F-scale estimate cannot be assigned. In- stead, for the historical dataset, the F-scale estimate is the minimum intensity that could cause the described or photographed damage. In this article, tornadoes without damage are unrated on the Fujita scale. Only 22% of the cases in the historical dataset were unrated, whereas 56% of the cases in the recent dataset were unrated. There is uncertainty in estimating the intensity of tornadoes using the F scale. Damage is not equivalent to intensity because the damage depends, not only upon the wind speed, but also upon the object receiving dam- age (Doswell and Burgess 1988; Doswell 2007). Terrain, building codes, debris, speed of the tornado, or rapidly fluctuating winds can contribute to the damage and af- fect the estimate of intensity. For example, tornadoes without damage are often classified as weak, although the true intensity could not be determined. Doswell and Burgess (1988) argued that many tornadoes have inap- propriate F ratings, perhaps by two categories or more. On the other hand, the F scale is largely based on damage to buildings, and the construction standards in Finland, as well as Europe as a whole, may differ from FIG. 1. (a) All and (b) significant tornado cases per decade in those in the United States (e.g., Feuerstein et al. 2011). Finland. The last decade includes only 2000–07.

3. Geographical distribution reports during the 1930s, 1990s, and 2000s. We hypoth- Tornado cases in our database are most frequent in esize that the movie Twister and an increase in television eastern and south-central Finland, as well as over the documentaries on tornadoes led to an increased aware- southern and southwestern seawaters (Fig. 2a). The oc- ness among the public about tornadoes in the late 1990s. currence is lowest in in northern Finland and Both the increased awareness and the onset of active in some inland areas of western Finland. The occurrence collecting of tornado reports show in the statistics, as of waterspout cases is high over the southern and south- during 1997–2007, a mean of 14 confirmed and proba- western seawaters, but also in the lakes of central Finland. ble tornado cases occurred each year, with a mean of If each waterspout were recorded separately instead of 11 tornado days. The tornado database is likely to be grouped into tornado cases, the frequency over the seas more consistent over time for the more intense torna- would be much larger. The large number of lakes and does (e.g., Brooks and Doswell 2001; Verbout et al. 2006). the large land area they cover helps to explain that 20% During 1930–2007, significant-tornado (F2 or stronger; of all tornado cases in this database spend part of their Hales 1988) cases occurred on average every second year, time over land and part of their time over water. but with a slight downward trend (Fig. 1b). The concentration of cases in eastern Finland is more evident when only the significant tornado cases are con- c. Estimating tornado intensity sidered (Fig. 2b). The annual probability of significant- The tornado intensity assessment is based on a dam- tornado cases was calculated during 1930–2007 because age survey, photographs, or eyewitness description of significant tornadoes are less sensitive to changing re- the damage. The estimate is based on the Fujita scale porting practice with time (e.g., Brooks and Doswell (Fujita 1981) and guidance tables for assigning tornado 2001; Verbout et al. 2006). The climatological probability damage to buildings (Minor et al. 1977, their Table 4; of at least one significant tornado within an 80 km 3 Bunting and Smith 1993, their appendix C). The esti- 80 km area during a year in several areas in central- mates were made by a single person (the first author), so southern and eastern-central Finland is 5%–7%. This the data should not contain some of the inhomogeneities means that an F2 or stronger tornado occurs in the area

Unauthenticated | Downloaded 10/01/21 07:20 AM UTC 1450 MONTHLY WEATHER REVIEW VOLUME 140

TABLE 2. Significant (F2 and greater) tornado cases in Finland.

Damage path Damage path No. of No. of Date Time (LT) Location width (m) length (km) Intensity fatalities injuries 1 Jul 1838 1600–1700 Turku 25–35 F2 27 May 1930 1240–1250 Paltamo 3–4 F2 29 Jun 1931 1600–1800 Vehmersalmi 60 2 F2 4 Aug 1932 1500 Nurmija¨rvi 400–500 14 F3 1 1 4 Aug 1932 1530 Hausja¨rvi 100–150 39 F3 6 Aug 1932 1700 , Nurmes 100–150 7 F2 6 Aug 1932 1915 40–50 4 F2 11 Sep 1932 1700 Kannus 100 10 F2 12 Jul 1933 1400 Toholampi 2000 6 F3 11 Jul 1934 1125 200 15 F4 1 11 Jul 1934 1330 Pulkkila 200 10 F3 1 9 Aug 1948 1200 F2 1 Oct 1949 1615 Joutsa 100–150 F2 29 Jul 1954 1200 , Vehkalahti, 20 F2 9 Jun 1956 Nilsia¨, F2 27 Jul 1957 1800–1900 Imatra 2500–3000 10 F2 3 27 Jul 1957 2000–2100 Luhanka 25–30 F2 27 Jul 1957 Karvia F2 10 Jul 1958 1500 Polvija¨rvi F2 1 Aug 1961 Loviisa 250 F2 1 Aug 1961 Orimattila, Hollola, Lammi F2 29 Jun 1967 Evening Vierema¨ 100 F2 26 Jul 1967 0830 Nurmo 1.5 F2 4 Aug 1967 Uusikaarlepyy 2000 7–8 F2 26 Aug 1967 1400 300 F2 7 Sep 1967 Evening Suomussalmi F2 8 Jul 1972 Evening Imatra, Joutseno, Puumala F2 1 l 12 Sep 1974 0900 Pyhta¨a¨, 50 F2 1 15 Aug 1975 1750 Orimattila, Iitti, 150 29 F2 1 30 Aug 1975 1035 50 1 F3 27 Jun 1976 1200 Leppa¨virta 500 2 F2 11 Aug 1985 1330 , 2000–3000 20 F2 7 Jul 1988 1350 Lieto 0.5 F2 22 Jul 1988 1630 Turku F2 3 Aug 1989 1800 Padasjoki 200 F2 28 Aug 1994 0600 150 0.35 F3 15 Jun 1995 1730 Na¨rpio F2 23 Jun 1997 1805 Lieksa 300 1.2 F2 12 Aug 1997 1800–1900 Kaustinen 150 1 F2 12 Jun 1998 1515–1540 Mikkeli 330 8.6 F2 15 Jul 2000 1800 Ja¨msa¨nkoski 50 15 F2 9 Aug 2001 1900 Rautavaara 150 1 F2 20 Aug 2004 1655–1720 Polvija¨rvi, Kontiolahti 20.2 F2

of maximum probability once every 14–20 yr. The belt cloud-to-ground lightning (e.g., Rauhala and Punkka of highest risk extends from the coast of the Gulf of 2008). In southern Finland, where the yearly mean cloud- Finland over central Finland to the Gulf of Bothnia. to-ground flash density is low (Tuomi and Ma¨kela¨ 2008; The geographical distribution of tornado cases in Ma¨kela¨ et al. 2011, their Fig. 1b), tornado cases and sig- Finland is similar to the geographic distribution of nificant tornado cases are relatively frequent. However, thunderstorm days (Fig. 2a, lightning data adapted from cloud-to-ground flash density is only one measure of Ma¨kela¨ et al. 2011, their Fig. 2b), although cloud-to- the location of deep moist convection. Although the basic ground lightning flash rate itself is not a good indicator conditions necessary for tornadoes and cloud-to-ground of tornado occurrence (e.g., MacGorman et al. 1989; lightning occurrence are not the same, both phenomena Perez et al. 1997; Teittinen and Ma¨kela¨ 2008). Indeed, require the existence of a convective cloud and a strong some tornadoes in Finland have been observed without updraft. Therefore, the geographical distribution of

Unauthenticated | Downloaded 10/01/21 07:20 AM UTC MAY 2012 R AUHALA ET AL. 1451

FIG. 2. Geographical distribution of (a) all tornado cases during 1796–2007 in Finland, plotted by the F scale as in the legend. (b) Annual probability (in percent) of at least one significant tornado in an 80 km 3 80 km area based on the 1930–2007 statistics. (c) Geographical distribution of severe hail cases in Finland during 1930–2006 from Tuovinen et al. (2009). All figures are overlaid on average annual number of thunderstorm days for Finland from Ma¨kela¨ et al. (2011). The units are thunderstorm days per year. thunderstorm days can be compared to the tornado oc- where only small areas of central Finland and northern currence. The frequency of thunderstorm days in Finland Finland have an elevation of more than 200 m above is highest in central Finland, where the yearly average is sea level, and tornado occurrence does not seem to be around 15 days (Fig. 2). This area coincides quite well related to elevation or gradients in elevation. The land– with the relatively frequent areas for significant tornado sea- and land–lake-induced boundaries near the coast cases. In Lapland, where the tornado cases are relatively and near lakes in eastern Finland may provide a favor- infrequent, 4–11 thunderstorm days occur per year. Thus, able environment for tornadogenesis, but that hypoth- the average number of thunderstorms days seems to esis is not explored within this article. correspond roughly to the significant tornado case fre- The small sample size of tornado cases in Finland at quency. any given location allows only an estimate of the broad Also, severe hail (Fig. 2c, hail data adapted from spatial patterns of tornado occurrence (e.g., Doswell Tuovinen et al. 2009) seems to have a similar geo- 2007). Also, because of the sparse population of Finland graphical distribution as tornado cases in Finland, al- (an average of 16 inhabitants per square kilometer, equal though a few of the cases in the two datasets are from the to that of Colorado), tornadoes are very likely under- same events. The main differences in the geographical reported and may be biased toward population centers. distributions are the larger number of severe hail cases Indeed, more tornado reports are received in areas of compared to tornado cases in agricultural areas in west- regionally high population density, especially for weak ern Finland, and the smaller number of severe hail cases tornadoes. In Lapland in the northern part of Finland, compared to tornado cases over coastal waters and in where the population density is much lower (two in- southeast Finland, where 30% of the surface are lakes. habitants per square kilometer, equal to that of Wyom- The likelihood of severe hail events being recorded over ing), the number of known tornado events is small, even agricultural areas and unlikelihood of them being re- though in southern and central Lapland there is an av- corded on water surfaces compared to tornadoes may erage of 8–11 thunderstorm days annually (Ma¨kela¨ et al. partly explain these differences. Despite these differ- 2011). ences, the similarity of geographical distributions of tornado cases and large hail motivates the question of 4. Intensity distribution whether geographically favored locations exist for the most severe convective weather in Finland. Most (86%) of the 298 tornado cases in our database Unfortunately, we have few hypotheses for why such were of F1 intensity or less (Fig. 3). A total of 43 sig- a distribution of tornadoes, if representative of a larger nificant tornado cases have been observed (Table 2). sample, would occur. An explanation from the orogra- There are differences in the intensity distribution be- phy is not apparent. Finland is a relatively flat country, tween the historical and recent datasets. Specifically,

Unauthenticated | Downloaded 10/01/21 07:20 AM UTC 1452 MONTHLY WEATHER REVIEW VOLUME 140

Finland. In Lapland and in large parts in western Finland, significant tornadoes have not been observed. However, because the F-scale rating depends on the damage the tornado causes, tornadoes in rural areas may be un- derrated (e.g., Doswell and Burgess 1988). Most tor- nadoes and waterspouts occurring at the coast were F1 or weaker. If all waterspouts were recorded as single events, the portion of nondamaging or weak tornado cases would be higher, especially in the recent dataset.

5. Monthly distribution FIG. 3. Number of tornado cases both in historical (1796–1996) and more recent (1997–2007) datasets by the F scale. Tornadoes have only been observed in Finland from April to November, with almost three quarters of tor- nado cases (73%) occurring during the warmest months 29% of the tornado cases from 1796–1996 were signifi- in July and August (Fig. 4a). In comparison, cloud-to- cant tornado cases, but only 4% of the tornado cases ground lightning may occur in Finland from April to from 1997–2007 were significant. The smaller percent- November (Tuomi and Ma¨kela¨ 2008), with the cloud-to- age of significant tornado cases in the recent dataset can ground flash rate the highest in continental Finland in be explained by the more efficient collection of tornado July and highest over the seas in July and August (Tuomi reports, especially for F1 and weaker tornadoes, similar and Ma¨kela¨ 2007). The majority (94%) of severe hail to that seen in the United States (Brooks and Doswell cases in Finland occur between June and August, with 2001; Verbout et al. 2006), and the fewer number of in- July being the peak month (Tuovinen et al. 2009). tense tornado cases in the shorter 1997–2007 period Based on the 1997–2007 statistics, the mean number of (Fig. 3). Because stronger tornadoes typically have larger tornado days by month is 1 in June, 5 in July, 4 in August, effects on society and influence larger areas (Brooks and 1 in September. Days with at least three tornadoes 2004), observations of the more intense tornadoes tend have been observed from July to October (Fig. 4b). For to be less sensitive to variations in reporting practice over all tornado cases that start on the land, the maximum is time (e.g., Verbout et al. 2006), explaining the large in July (Fig. 4c). Of all tornado cases in July, more than proportion of significant tornado cases in the historical half (54%) are weak and start on land. The maximum for dataset. significant tornadoes is in August when 17% of all ob- There have been only six F3 cases, one F4, and no F5s served tornado cases are significant (Fig. 4d). Days with recorded in our database. The F4 tornado occurred on at least two significant tornado cases have been observed 11 July 1934 at Kiuruvesi in central Finland. By compar- only in July and August. ison, F4 tornadoes have been observed in several other Waterspout cases tend to occur toward late summer European countries: Estonia (Tooming 2002), Germany compared to all tornado cases (Fig. 4c). Therefore, if (Dotzek 2001), and the Czech Republic (Setva´ketal. each waterspout were recorded as a single case instead 2003). In France, a few F5 tornadoes have been ob- of grouping them into tornado cases, the monthly dis- served (Paul 2001). As in Finland (Fig. 3), a relatively tribution of tornado cases in Finland would shift more small number of tornadoes in these countries are strong. toward late summer and early autumn. The maximum The decline with intensity is similar to that seen in number of waterspout cases in our database occurs in Brooks and Doswell (2001) and Dotzek et al. (2003). August when 53% of the tornado cases start over water, When greater attention is put into collecting reports although the maximum percentage of waterspout cases or there is greater public awareness of tornadoes, as in occurs in October when 83% of the tornado cases start Ireland (Tyrrell 2003), the relationship is closer to being over water. In July, the peak month in Finland for tor- log–linear. The impact of the onset of active collecting nado cases, only 36% of tornado cases are waterspouts. of reports can be also seen in Finland, where the number There are only a few known waterspout cases in May, of weak tornado cases has increased in the recent da- June, October, and November. This is consistent with taset compared to the historical dataset (Fig. 3). This the cold sea surface temperatures in the early summer underreporting of weaker tornadoes has been observed being less favorable for convection than the warmer in other countries (e.g., Germany, France, etc.). temperatures that occur in the late summer. In east-central Finland, the fraction of tornado cases Some minor differences in the monthly distribution that were significant was higher than elsewhere in between different geographical locations can be found

Unauthenticated | Downloaded 10/01/21 07:20 AM UTC MAY 2012 R AUHALA ET AL. 1453

FIG. 4. (a) Monthly distribution of all tornado cases in Finland in 1796–2007. (b) Distribution of days with at least three tornado cases. (c) Distribution of tornado cases over land and water surface. (d) Monthly distribution of significant (F21) tornado cases. in our database, although the small sample size affects 0900–2059 LT (Fig. 5a). There were only 10 tornado the credibility of the interpretation (e.g., Doswell 2007). cases at night (between 2100 and 0659 LT), and no cases In western inland areas, tornado cases seem to occur between 0100 and 0459 LT. The diurnal distribution of mostly in July and August; in the east, they occur during waterspout cases was more evenly spread throughout the whole season. Most tornado cases offshore occur in the day than tornado cases over land (Fig. 5b). Most August or September, and only a few rare events are (69%) of the tornado cases over land occurred in the known to occur in June. late afternoon and evening, between 1500 and 2059 LT, As in Finland, tornadoes in neighboring Estonia with a peak at 1700–1859 LT, whereas waterspout cases (Tarand 1995) and Sweden (Peterson 1998) occur most peaked at noon (Fig. 5b). If single waterspouts were frequently during July and August. In central European recorded separately instead of grouping them into countries, the peak occurs earlier in the summer: in tornado cases, the diurnal maximum of tornadoes in Germany and in Austria in July (Dotzek 2001; Holzer Finland would have two peaks, one before noon and 2001) and in the Czech Republic and Hungary between one in the afternoon. May and July (Setva´k et al. 2003; Szilard 2007). In Med- The afternoon maximum in tornado cases is consistent iterranean countries, the peak is in late summer or au- with the diurnal cycle of thunderstorms in Finland, al- tumn (e.g., Paul 2001; Gaya´ et al. 2001; Sioutas 2003, 2011; though the diurnal maximum in cloud-to-ground flash Giaiotti et al. 2007; Gaya´ 2011). Fujita (1973) found that rate in mainland Finland typically peaks around 1500– July and August are the peak months for tornadoes from 1659 LT (Tuomi and Ma¨kela¨ 2008), two hours before northern Russia to Germany, France, and northern Italy, the maximum of tornado cases. In comparison, large hail whereas, in other parts of western and southern Europe, (2–3.9 cm in diameter) is most frequent between 1400 the peak moves toward the autumn. In the United States, and 1800 LT and very large hail (4 cm or larger) between the peak frequency for tornadoes is in May and June 1600 and 2000 LT (Tuovinen et al. 2009). Tuovinen et al. (Verbout et al. 2006). (2009) speculated this shift to be related to delayed ini- tiation of convection in slightly capped environments, which favor the most intense updrafts. We speculate 6. Diurnal distribution that capping inversions may occur in some of the tornado The tornado cases that had sufficient information on situations in Finland, which would explain the later max- time of occurrence were placed in 2-h bins in local imum of tornado cases compared to the cloud-to-ground standard time (LT); 0500–0659, 0700–0859 LT, etc. Most flash-rate maximum. Similar diurnal distributions of tor- of the tornado cases in our database occurred between nadoes as in Finland occur in France (Dessens and Snow

Unauthenticated | Downloaded 10/01/21 07:20 AM UTC 1454 MONTHLY WEATHER REVIEW VOLUME 140

7. Conclusions This study summarizes the tornado climatology of Finland derived from 298 observed tornado cases during 1796–2007. The reports are collected and evaluated, re- sulting in two datasets: a historical dataset from 1796– 1996 with 129 cases and a recent dataset from 1997–2007 with 169 cases. The number of tornado cases in our da- tabase is highest from the area extending from southern and southwestern seawaters to eastern Finland. The re- gion of highest frequency of significant tornado cases extends from the Gulf of Finland to the Gulf of Bothnia. The geographical distribution of weak tornado cases shows higher concentrations of reports near densely populated areas, suggesting population bias and underreporting of weak tornadoes in sparsely populated areas. Likely underreporting of weak tornadoes can also be seen in the historical dataset when comparing the in- tensity distribution of the historical and recent datasets. Whereas in the historical dataset, 29% of tornado cases were significant, in the recent dataset only 4% were sig- nificant. In the recent dataset, the large number of weak FIG. 5. Diurnal distribution of (a) all tornado cases and signifi- tornado cases is the result of an efficient collection of cant tornado cases and (b) tornado cases over water and land reports, whereas significant tornadoes are more likely surface in Finland. One column indicates a 2-h period 0500–0659, 0700–0859 LT, etc. There are no tornado cases recorded between to be independent of the efforts put into collecting re- 0100 and 0459 LT. ports. Altogether, 14% (43 cases) of all recorded tornado cases in Finland are significant. The strongest docu- mented tornado is of F4 intensity. In our database, an average of 14 tornado cases occurs 1989), the United States (Kelly et al. 1978), and Germany every year in Finland from April to November. Coin- (Dotzek 2001), for example. cident with the highest lightning occurrence both over The observed diurnal cycle of waterspout cases in land and sea, the peak months for tornado cases are July Finland with a noon peak is similar to that observed by and August. August is the peak month for both water- Dotzek et al. (2010) with waterspouts near the German spout cases (53% of all August tornado cases) and sig- coast at the North Sea and Baltic Sea. Distributions nificant tornado cases (17%), whereas typically (52%) with three maxima—two higher peaks near noon and in July tornado cases are weak tornado cases over land. the late afternoon, and a lower peak in morning—have Most tornado cases over land occur between 1500 and been observed by Golden (1973) with waterspouts in 2059 LT, whereas tornado cases over water tend to occur the Lower Florida Keys and by Sioutas (2011) in Greece more equally throughout the day. (both tornadoes and waterspouts). A possible explana- tion for the more evenly spread diurnal distribution of tornado cases over water compared to tornado cases Acknowledgments. We thank Sylvain Joffre (Finnish over land is that the warm water surface may be a fa- Meteorological Institute) for his comments on an earlier vorable location for the development of convection at version of the manuscript and Pentti Pirinen (Finnish any time of the day or night. Accordingly, the cloud-to- Meteorological Institute) for providing data for Fig. 2b. ground flash rate over the sea is somewhat more evenly Partial funding for Schultz comes from Vaisala Oyj and distributed throughout the day than over land areas, Grant 126853 from the Academy of Finland. but there is still a distinct afternoon maximum and minor morning maximum (Tuomi and Ma¨kela¨ 2008). Although the diurnal cycle of thunderstorms over water likely has REFERENCES a large influence on waterspout occurrence, the lack of Angervo, J. M., 1949: Pyo¨rremyrskyt lokakuun 1. P:na¨ 1949 (Tor- tornado reports at night may also be because of darkness nadoes on 1st October 1949.) IK, kuukausikatsaus, 43 (10),p.4. (in the late summer) or the smaller number of people Brooks, H. E., 2004: On the relationship of tornado path length and outdoors. width to intensity. Wea. Forecasting, 19, 310–319.

Unauthenticated | Downloaded 10/01/21 07:20 AM UTC MAY 2012 R AUHALA ET AL. 1455

——, and C. A. Doswell III, 2001: Some aspects of the international Severe Local Storms, Baltimore, MD, Amer. Meteor. Soc., climatology of tornadoes by damage classification. Atmos. 165–168. Res., 56, 191–201. Holden, J., and A. Wright, 2004: UK tornado climatology and the ——, and N. Dotzek, 2008: The spatial distribution of severe con- development of simple prediction tools. Quart. J. Roy. Meteor. vective storms and an analysis of their secular changes. Cli- Soc., 130, 1009–1021. mate Extremes and Society, H. F. Diaz and R. Murnane, Eds., Holzer, A. M., 2001: Tornado climatology of Austria. Atmos. Res., Cambridge University Press, 35–53. 56, 203–211. Bunting, W. F., and B. E. Smith, 1993: A guide for conducting Ja¨rvi, L., and Coauthors, 2007: Micrometeorological observations convective windstorm surveys. NOAA Tech. Memo. NWS of a microburst in southern Finland. Bound.-Layer Meteor., SR146, Scientific Services Division, Southern Region, Fort 125, 343–359. Worth, TX, 44 pp. Kelly, D. L., J. T. Schaefer, R. P. McNulty, and C. A. Doswell III, Dessens, J., and J. T. Snow, 1989: Tornadoes in France. Wea. 1978: An augmented tornado climatology. Mon. Wea. Rev., Forecasting, 4, 110–132. 106, 1172–1183. ——, and ——, 1993: Comparative description of tornadoes in Kera¨nen, J., 1930: Voimakas pyo¨ rremyrsky Paltamon Mieslahdella France and the United States. The Tornado: Its Structure, Dy- toukok. 27. P:na¨ 1930 (A strong tornado at Mieslahti in Pal- namics, Prediction and Hazards, Geophys. Monogr., Vol. 79, tamo on 27th May 1930.) IK, kuukausikatsaus, 24 (11), p. 4. Amer. Geophys. Union, 427–434. Leitao, P., 2003: Tornadoes in Portugal. Atmos. Res., 67–68, 381– Doswell, C. A., III, 2007: Small sample size and data quality issues 390. illustrated using tornado occurrence data. Electron. J. Severe MacGorman, D. R., D. W. Burgess, V. Mazur, W. D. Rust, W. L. Storms Meteor., 2 (5), 1–16. Taylor, and B. C. Johnson, 1989: Lightning rates relative to ——, and D. W. Burgess, 1988: On some issues of U.S. tornado tornadic storm evolution on 22 May 1981. J. Atmos. Sci., 46, climatology. Mon. Wea. Rev., 116, 495–501. 221–251. ——, H. E. Brooks, and N. Dotzek, 2009: On the implementation Ma¨kela¨, A., P. Rossi, and D. M. Schultz, 2011: The daily cloud-to- of the enhanced Fujita scale in the USA. Atmos. Res., 93, 554– ground lightning flash density in the contiguous United States 563. and Finland. Mon. Wea. Rev., 139, 1323–1337. Dotzek, N., 2001: Tornadoes in Germany. Atmos. Res., 56, 233– Marcinoniene, I., 2003: Tornadoes in Lithuania in the period of 251. 1950–2002 including analysis of the strongest tornado of 29 ——, J. Grieser, and H. E. Brooks, 2003: Statistical modeling of May 1981. Atmos. Res., 67–68, 475–484. tornado intensity distributions. Atmos. Res., 67, 163–187. Minor, J. E., J. R. McDonald, and K. C. Mehta, 1977: The tornado: ——, P. Groenemeijer, B. Feuerstein, and A. M. Holzer, 2009: An engineering-oriented perspective. NOAA Tech. Memo. Overview of ESSL’s severe convective storms research using ERL NSSL-82, National Weather Service, Norman, OK, the European Severe Weather Database ESWD. Atmos. Res., 196 pp. 93, 575–586. Newark, M. J., 1984: Canadian tornadoes, 1950–1979. Atmos.– ——, S. Emeis, C. Lefebvre, and J. Gerpott, 2010: Waterspouts Ocean, 22, 343–353. over the North and Baltic Seas: Observations and climatology, Outinen, K., and J. Teittinen, 2008: Polarimetric radar observations predictability and reporting. Meteor. Z., 19, 115–129. of a tornadic supercell in Finland. Proc. Fifth European Conf. Elsom, D. M., and G. T. Meaden, 1982: Tornadoes in the United on Radar in Meteorology and Hydrology, Helsinki, Finland, Kingdom. Preprints, 12th Conf. on Severe Local Storms, San Finnish Meteorological Institute, P4.3. [Available online at Antonio, TX, Amer. Meteor. Soc., 55–58. http://erad2008.fmi.fi/proceedings/extended/erad2008-0064- Feuerstein, B., P. Groenemeijer, E. Dirksen, M. Hubrig, A. M. extended.pdf.] Holzer, and N. Dotzek, 2011: Towards an improved wind Paul, F., 2001: A developing inventory of tornadoes in France. speed scale and damage description adapted for central Atmos. Res., 56, 269–280. Europe. Atmos. Res., 100, 547–564. Perez, A. H., L. J. Wicker, and R. E. Orville, 1997: Characteristics Forbes, G. S., and R. M. Wakimoto, 1983: A concentrated outbreak of cloud-to-ground lightning associated with violent tornadoes. of tornadoes, downbursts and microbursts, and implications Wea. Forecasting, 12, 428–437. regarding vortex classification. Mon. Wea. Rev., 111, 220–236. Peterson, R. E., 1982: Tornadic activity in Europe the last half- Fujita, T. T., 1973: Tornadoes around the world. Weatherwise, 26, century. Preprints, 12th Conf. on Severe Local Storms, San 56–62. Antonio, TX, Amer. Meteor. Soc., 63–66. ——, 1981: Tornadoes and downbursts in the context of general- ——, 1992: Letzmann and Koschmieder’s ‘‘Guidelines for research ized planetary scales. J. Atmos. Sci., 38, 1511–1534. on funnels, tornadoes, waterspouts and whirlwinds.’’ Bull. Gaya´, M., 2011: Tornadoes and severe storms in Spain. Atmos. Amer. Meteor. Soc., 73, 597–611. Res., 100, 334–343. ——, 1995: Johannes Peter Letzmann: Pioneer tornado researcher. ——, V. Homar, R. Romero, and C. Ramis, 2001: Tornadoes and Meteorology in Estonia in Johannes Letzmann’s Times and waterspouts in the Balearic Islands: Phenomena and environ- Today, H. Eelsalu and H. Tooming, Eds., Estonian Academy ment characterization. Atmos. Res., 56, 253–267. Publishers, 9–43. Giaiotti, D. B., A. Pucillo, and F. Stel, 2007: The climatology of ——, 1998: Tornadoes in Sweden. Preprints, 19th Conf. on Severe tornadoes and waterspouts in Italy. Atmos. Res., 83, 534–541. Local Storms, Minneapolis, MN, Amer. Meteor. Soc., 89–92. Glickman, T. S., Ed., 2000: Glossary of Meteorology. 2nd ed. Amer. ——, 2000: Tornadoes of the former Soviet Union. Preprints, 20th Meteor. Soc., 855 pp. Conf. on Severe Local Storms, Orlando, FL, Amer. Meteor. Golden, J. H., 1973: Some statistical aspects of waterspout forma- Soc., 138–141. tion. Weatherwise, 26, 108–117. Punkka, A.-J., and M. Bister, 2005: Occurrence of summertime Hales, J. E., Jr., 1988: Improving the watch/warning program convective precipitation and mesoscale convective systems in through use of significant event data. Preprints, 15th Conf. on Finland during 2000–01. Mon. Wea. Rev., 133, 362–373.

Unauthenticated | Downloaded 10/01/21 07:20 AM UTC 1456 MONTHLY WEATHER REVIEW VOLUME 140

——, J. Teittinen, and R. H. Johns, 2006: Synoptic and mesoscale Teittinen, J., and A. Ma¨kela¨, 2008: Lightning and radar reflectivity analysis of a high-latitude derecho–severe thunderstorm out- signatures in tornadic supercell thunderstorms in Finland. break in Finland on 5 July 2002. Wea. Forecasting, 21, 752–763. Proc. Fifth European Conf. on Radar in Meteorology and Hy- Rauhala, J., and A.-J. Punkka, 2008: Radar observations of a tor- drology, Helsinki, Finland, Finnish Meteorological Institute, nadic severe frontal rainband. Preprints, 24th Conf. on Severe 4 pp. [Available online at http://erad2008.fmi.fi/proceedings/ Local Storms, Savannah, GA, Amer. Meteor. Soc., P9.6. [Avail- extended/erad2008-0129-extended.pdf.] able online at http://ams.confex.com/ams/24SLS/techprogram/ ——, H. Hohti, J. LaDue, and R. A. Brown, 2006: Analysis of paper_142151.htm.] a tornadic mini-supercell in Finland by using Doppler radar. ——, and D. M. Schultz, 2009: Severe thunderstorm and tornado Preprints, 23rd Conf. on Severe Local Storms, St. Louis, MO, warnings in Europe. Atmos. Res., 93, 369–380. Amer. Meteor. Soc., P6.1. [Available online at ams.confex. Reynolds, D. J., 1999: Revised UK tornado climatology, 1960– com/ams/pdfpapers/115430.pdf.] 1989. J. Meteor. UK, 24, 290–321. Tooming, H., 2002: Strong tornadoes in Estonia. Extended Ab- Rossi, V., and M. Franssila, 1960: Ukkonen ja pyo¨ rremyrskyt stracts, European Conf. on Severe Storms 2002, Prague, Czech (Thunderstorm and tornadoes). Oma maa, 8, 54–65. Republic, Czech Hydrometeorological Institute, p. 33. Setva´k, M., M. Salek, and J. Munzar, 2003: Tornadoes within the ——, H. Kotli, and R. E. Peterson, 1995: Vigorous tornadoes and Czech Republic: From early medieval chronicles to the ‘‘in- waterspouts during the last 35 years in Estonia. Meteorology in ternet society.’’ Atmos. Res., 67–68, 589–605. Estonia in Johannes Letzmann’s Times and Today, H. Eelsalu Simojoki, H., 1931: Pyo¨rremyrsky Wehmersalmella kesa¨k. 29. p. and H. Tooming, Eds., Estonian Academy Publishers, 168– 1931 (Tornado at Wehmersalmi on 29 June 1931). IK, 179. kuukausikatsaus, 25 (10),p.4. Tuomi, T., and A. Ma¨kela¨, 2007: Lightning observations in Finland, ——, 1935: Die Tromben am 6. August 1932 im inneren Finnland 2007. Finnish Meteorological Institute, Rep. 2007:5, 49 pp. (Tornado on 6 August 1932 in central Finland). Ann. Acad. ——, and ——, 2008: Thunderstorm of Finland 1998–2007. Sci. Fenn., 44A, 1–26. Geophysica, 44 (1–2), 67–80. Sioutas, M. V., 2003: Tornadoes and waterspouts in Greece. Atmos. Tuovinen, J.-P., A.-J. Punkka, J. Rauhala, H. Hohti, and D. M. Res., 67–68, 645–656. Schultz, 2009: Climatology of severe hail in Finland: 1930– ——, 2011: A tornado and waterspout climatology for Greece. 2006. Mon. Wea. Rev., 137, 2238–2249. Atmos. Res., 100, 344–356. Tyrrell, J., 2003: A tornado climatology for Ireland. Atmos. Res., Szilard, S., 2007: A systematic approach to synoptic tornado cli- 67–68, 671–684. matology of Hungary for the recent years (1996–2001) based Verbout, S. M., H. E. Brooks, L. M. Leslie, and D. M. Schultz, 2006: on official damage reports. Atmos. Res., 83, 263–271. Evolution of the U.S. tornado database: 1954–2004. Wea. Tarand, A., 1995: How often do tornadoes occur in Estonia? Me- Forecasting, 21, 86–93. teorology in Estonia in Johannes Letzmann’s Times and To- Wegener, A., 1917: Wind- und Wasserhosen in Europa (Torna- day, H. Eelsalu and H. Tooming, Eds., Estonian Academy does and Waterspouts in Europe.) Fredr. Vieweg & Sohn, Publishers, 132–138. 301 pp.

Unauthenticated | Downloaded 10/01/21 07:20 AM UTC